Beruflich Dokumente
Kultur Dokumente
244
A. P. de Leon
Fischer suggests the availability of different design rationalisation strategies for taking advantage of the fabrication economies gained with massproduction, as for example the rationalisation strategy employed in the Beijing
National Swimming Centre allowed a reduction of the number of prefabricated unique building components in the assembly and simultaneously met
the aesthetic goals of apparent irregularity (Fischer 2007). Fischer resurfaces
three strategic opportunities in the rationalisation of unconventional forms
originally presented by Hugh Whitehead, of the foster and partners Specialist Modelling Group; the three strategies deal with firstly the opportunities
in decision making (pre-rationalisation), secondly how to cope with design
consequences (post-rationalisation) and thirdly a hybrid strategy namely corationalisation where parallel decisions affecting the rationalisation of form
can be made alongside the process of design.
This paper will demonstrate a faade rationalisation strategy developed at
Gehry Technologies for mitigating the tradeoffs between project constructability and the aesthetic implications of arranging flat hexagonal panels over the
unconventional geometry of the Museo Soumaya in visually pleasing arrangements. The rationalisation strategy had to manage the tradeoffs between a low
number of unique building components and the effects these have over the
perceivable irregularities of the panel-to-panel gaps (Fischer 2007).
2. Initial assumptions and rules
The faade rationalisation strategy of the Museo Soumaya was guided by a
finite set of rules imposed by the design team as follows: 1) maintain a uniform
gap between all six sides of the hexagon panel system; 2) begin with a standard hexagon size of dimension 63 cm diameter; 3) force the hexagon system to
grow in scale (from a diameter range of 63 cm to 175 cm) as the pattern turns
the corners of the faade; 4) group the panels into families, by minimising the
amount of unique panels that need to be fabricated.
Parting from these rules, an addition of two important tactical design decisions guided the process of rationalisation, the first of these decisions opted in
selecting early in the project a pre-rationalised turn-key solution patented by
Geometrica for the fabrication of freeform space frame structures, the system
would serve both as the panel positioning device and the support structure for
both the aluminium panels and the waterproofing panels. The second decision
was to freeze the MDS Master Design Surface as a fixed design component
in the project; this is due to concurrent assembly processes within the interior
of the project completing at a faster pace than that of the exterior faade engineering exercises, in order to avoid errors in misalignments further down the
245
construction process, the MDS was frozen, dimensions can be extracted from
slicing and dicing the surface, although the MDS cannot be edited further.
As a design consequence of selecting Geometricas space frame strut and
node system, the rationalisation strategy was reduced substantially, triangular
configurations of same size circles produce hexagonal intersection patterns,
each node in Geometricas space frame holds the centre of gravity of a single
hexagonal panel and three incoming struts, forming a secondary triangular
aluminium structure over the primary steel of the project providing support for
hanging the outer aluminium panels and the inner waterproofing panels also
developed by Geometrica.
The information requirements needed by Geometrica for the assembly of
the secondary aluminium space frame using their proprietary system needed
only the XYZ position and XYZ normals of each node in the assembly, as a
result of a one to one relationship between the node and centre of gravity of
the panels, obtaining a packing of panels over the surface and extracting the
centre of gravity of the packing, provides sufficient information for coordinating the overall panel, strut and node assemblies.
In the following sections two unsuccessful rationalisation methods for
obtaining the coordination data are examined, while the methods arrive at
hexagonal packings of panels over the MDS, they fail to address the design
rules outlined in Section 2.
3. Conformal mappings
An often employed rationalisation strategy for obtaining panels over a
doubly curved MDS is to roughly speaking wrap a 2D pattern drawing over
the surface. In mathematical terms this operation is known as a conformal
mapping. Depending on the results obtained, a mapping could exhibit two
useful properties, firstly if the lengths in the original 2D drawing are preserved
after transformation, the mapping is said to be isometric, and secondly if the
angles between lines in the original 2D drawing are preserved after transformation, the mapping is said to be conformal.
The optimal solution to the rationalisation strategy would be a mapping
that exhibits both isometric and conformal properties. In the case of the Museo
Soumaya the MDS was defined as a NURBS surface digitised from a physical model. When wrapping 2D drawings over a NURBS surface, the mapping
is neither length preserving nor angle preserving, unless the MDS is either a
cylinder or a plane, in which case there is still the possibility of deformations
in the conformal mappings due to the spacing of surface knots not being uniformly distributed over the surface.
246
A. P. de Leon
4. Curvature gradients
Another traditional useful rationalisation strategy is to map the local Gaussian
curvature properties onto the MDS, and from the mapping extracting the areas
in the surface with relatively low curvature conditions, developable areas.
Even though this approach allows wrapping 2D patterns over the Developable
regions of the surface, the method cannot handle easily the transition between
one area and its adjacent neighbour. Subdivision of the MDS by the Gaussian Curvature coefficient can prove to be useful in other panelling situations
where a seamless panel distribution is not such a high aesthetical priority.
5. Circle hex meshes/sphere packing
This section presents the final successful rationalisation strategy employed for
obtaining the panel, strut and node assembly coordination data for manufacturing and assembly. Intersecting circles produce the best hexagonal distributions
in two dimensional spaces as shown in Figure 1, there is a one to one relation
between the hexagon diameters and the circle diameters that produce them.
To control the diameter and the gap of the hexagon, we build a circle mesh of
intersected circles, constructing the diameters of the circles from the values of
the desired diameter plus the value of the desired gap, we suspend the actual
hexagon by inscribing it on a circle offset inward using 1/2 of the desired
gap as the offset value, a diagram of the extra wide circle and the suspended
hexagon is shown on Figure 2.
247
Following Thompsons insights of the polar furrow, we can construct any semihexagon as a local property of the intersection of two spheres with a doubly
curved surface and then intersecting the two resultant curved circles supported
on the surface and subsequently getting the north and south point, if we choose
only the south point to locate another sphere and repeat this process until we
have reached the end of the row, and then repeat the same process over and over
again obtaining a new row for each pass, we can obtain a circle mesh grid over
the surface as many rows and columns as needed see Figure 4 for the algorithmic process that creates the mesh, see Figure 4 for the result of the algorithm.
248
A. P. de Leon
From the circle mesh grid as pointed in Figure 4, the hexagonal pattern is
extracted by hanging hexagons inscribed in an offset circle 1/2 of the desired
gap.
6. Correcting stretching artefacts
Although the mesh is close to the desired result, the hexagonal mesh contracted
vertically on the regions of high curvature and as a result the outer edge rose
and retracted from the initial position due to the reduction in the height of the
panels. In order to correct these artefacts, a post-production process must be
executed over the resultant mesh to correct the height reduction; this process
corrects the height of each panel from its current reduced height to the height
of an ideal hexagon, which is a hexagon where each of its vertices has an
interior angle of 60 degrees. The reduced height panels selected are shown on
the left on Figure 5, and the result of correcting the panel heights is shown on
the right in Figure 5.
The process of correcting the heights of the panels requires first to identify all
the panels that require correction. Passing a curve vertically through all the
points in a column of panels, these curves are placed on top of the surface of
the envelope and extended past the boundary edge of the surface, effectively
these curves where used as rails to scroll the surface downwards until the vertical dimension of all hexagons corresponds to the horizontal dimension, there
is a height to width proportion on a 60 degree hexagon, this ratio can be used
to calculate the percentage needed to correct the vertical dimension of each
hexagon. Before this process is executed, the panels that exhibit an abnormal squashing are filtered and selected as shown in Figure 6, the panels with
abnormalities are referred to as custom panels and the ones with acceptable
conditions are referred to as standard, Figure 7 shows the process of extracting
rails from the panels and then stretching the panels over the extracted rails to
correct their abnormality.
249
The sphere packing and the hexagonal mesh extracted from it are sensitive
to initial conditions, after conducting many trial and error studies, the pattern
with the best aesthetics results were achieved by starting the sphere packing
horizontally in the middle of the envelope. The location of the first sphere in
the sphere packing algorithm has a high impact on the aesthetic output of the
overall mesh, we found the most homogenous solutions where achieved by
starting the sphere packing at a location on the surface with the least Gaussian
curvature, to identify this location, an optimisation process was executed to
identify this location, as shown in Figure 8.
Figure 8. Optimisation process for finding the starting point of the pattern.
250
A. P. de Leon
251
8. Conclusions
This report outlines various rationalisation strategies for negotiating tradeoffs
between geometrical and aesthetical implications.
The most successful rationalisation strategy found the most aesthetically
pleasing distribution of panels by using a stable sphere packing method, and
a self-adjusting pattern stretching algorithm, by extending Digital Project
through bespoke software solutions.
A solution to cataloguing the resultant panels was found using a straightforward approach involving the use of a k-means algorithm for the clustering process in Excel and the subsequent mapping process for returning the
result into Digital Project, by using a k-means algorithm as opposed to a
hierarchical agglomerative clustering algorithm, an approximate solution was
found at a faster time, while giving the user control to specify directly the
exact number of families desired.
252
A. P. de Leon
Acknowledgements
This work was developed during my employment at Gehry Technologies Inc, in collaboration with Jason Sidelko, Lee Waren, Eddy Slim and Rolando Mendoza. The writing efforts
are funded by the ARC Discovery grant Challenging the Inflexibility of the Flexible Digital
Model, I am grateful to the valuable feedback and comments of my PhD supervisors Mark
Burry and Jane Burry.
References
Fischer, T.: 2007, Rationalising bubble trusses for batch production, Automation in Construction, 16(1), 4553.
Glymph, J., Shelden, D., Ceccato, C., Mussel, J. and Schober, H.: 2002, A parametric strategy
for freeform glass structures using quadrilateral planar facets, in Proctor, G. (ed.), Proc.
22nd ACADIA Conference, Pomona, 307325.
Shelden, R. D.: 2002, Digital Surface Representation and the Constructability of Gehrys Architecture, PhD thesis, Massachusetts Institute of Technology, Cambridge.
Thompson, D.: 1961, On Growth and Form, Cambridge University Press, Cambridge.